The Logician.AML Watcher · Case study
Built the product function and helped compliance teams find relevant risks with better name matching, configurable scoring and context-aware media screening.
When I joined Programmers Force, AML Watcher had technology but no product organisation: no product team, no product function, no structured process. Engineering operated without PRDs. Sales operated without positioning materials. I was the first product hire. What existed was a compliance platform covering sanctions across 215+ regimes, PEP screening across 235+ countries, and watchlist screening against 3,500+ databases, but without the product function to scale it.
The data layer of 15 to 20 million entity profiles lacked consistent structure. Adverse media relied on basic keyword matching, producing massive false positive volumes. There was no risk scoring, no biometric capability, and no configurable client controls. The technology was real but the product was incomplete.
I built the product function from scratch: established the team, defined the processes, and created every operational artefact including PRDs, sprint planning cadence, pricing strategy, data dictionaries, API documentation, and all client-facing collateral. I worked directly with Engineering on database schemas and API contracts, with Data Science on matching algorithms and NLP pipelines, with Design on every user flow, and with Sales to ensure the narrative matched the reality.
The approach was simultaneously top-down (product strategy, pricing, competitive positioning) and bottom-up (restructuring MongoDB schemas, defining key-value pairs, running data quality checks in Python). In a startup, the PM does not choose between strategy and execution. You own both.
Took the system from keyword matching to a context-aware, sentiment-analysed, multilingual monitoring engine. The platform uses NLP and entity extraction to understand the role of each individual mentioned in media coverage, rather than simply flagging keyword hits. It monitors 5,000+ sources across 80+ languages, categorised into 415+ risk categories. This technology was later spun off as Media Watcher.
Explore the featureDesigned from scratch with a three-parameter model. Country Risk draws from FATF grey/black list status and the Corruption Perception Index. Database Risk accounts for PEP level, sanctions status, and watchlist recency. Criminal Activity Risk is derived from adverse media sentiment scores. Clients can adjust risk weights, set custom thresholds, and define risk appetite per jurisdiction. The engine achieved approximately 40% false positive reduction and a 70% decrease in processing time.
Explore the featureDefined matching logic across multiple dimensions including phonetic matching (Mohammad/Muhamad/Mohamed), diacritics handling, Arabic-to-Latin transliteration, prefix/suffix parsing (Al-, Bin, Abu), and alias/AKA detection. Reduced false positives by up to 60% across 15 to 20 million entity profiles.
Introduced facial matching into AML screening: 1:1 facial image matching against AML databases. Achieved a 44% false positive reduction and up to 90% fewer manual reviews. Designed to give reviewers another signal alongside name matching.
Explore the featureCrypto Wallet Screening, Custom Search Profiles, Case Management, Batch Search, Ongoing Monitoring, Transaction Monitoring (150+ AML typologies, 10,000+ custom rules), Complete API Documentation (doc.amlwatcher.com), and Custom Whitelisting/Blacklisting.
Built the product team from 2 to 10, establishing the product function from zero
40% false positive reduction and 70% processing time decrease via risk scoring
60% false positive reduction through fuzzy name matching
44% false positive reduction and 90% fewer manual reviews through biometric AML
57% monthly revenue growth sustained over a quarter through targeted feature enhancements
30% increase in enterprise sales through technical CTO/CEO demos
Adverse media technology scaled into a standalone product (Media Watcher)
AML Watcher taught me what it means to build from zero. Not just a feature, but an entire product organisation, including the team, the processes, the strategy, and the commercial outcomes. The most important decision was investing early in data architecture: restructuring the MongoDB schemas and building entity resolution logic before shipping new features. That foundational work made everything else possible at scale.